Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 12 Oct 1998 13:36:19 -0600 (MDT)
From:      "Kenneth D. Merry" <ken@plutotech.com>
To:        dillon@apollo.backplane.com (Matthew Dillon)
Cc:        mike@smith.net.au, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG
Subject:   Re: cvs commit: src/sys/cam/scsi scsi_da.c
Message-ID:  <199810121936.NAA26254@panzer.plutotech.com>
In-Reply-To: <199810121927.MAA16373@apollo.backplane.com> from Matthew Dillon at "Oct 12, 98 12:27:30 pm"

next in thread | previous in thread | raw e-mail | index | archive | help
Matthew Dillon wrote...
>     cache-sync is such a fundamental command, I can't imagine any drive
>     not doing the right thing with it, especially drives which support tagged
>     commands.  The command is tightly integrated into the disk cache paradigm
>     as given by the SCSI-II standard.
> 
>     But I guess there could be a few...  even so, I think we have to support
>     it.  It's just too important a command, especially in larger RAID SCSI 
>     systems which have huge write caches.  Consider a UPS signalling a machine 
>     that it's about to die and the machine trying to flush everything out and 
>     shut down.  Better to get the cache flushed to disk (non-volatile or not) 
>     before power goes poof.

Good example.

>     Have there been any specific reports of problems with the use of
>     the command other then the vague micropolis report in the first
>     email (first email that I got, anyway) ?

Well, basically Tom Jackson <toj@gorilla.net> has three Fujitsu drives, two
of them are the M2954's that are quirked in the da driver now.  He sent one
of those drives to Justin this summer, and we figured out that it was
returning the cache sync command with an illegal request sense key.  That
isn't really a big problem, since we just ignore the error.  It looks like
his other drive (slightly different version of the same model) hangs when
it gets the cache sync command.  I'm not sure that's what is happening, but
it seems to be the case.

Eventually, I think, the command timeout should hit and the machine should
reboot, though.

The reason he didn't see this before my most recent changes is because the
drive in question is his boot drive, and for some reason wasn't getting
final close.  The shutdown hook I put in does a cache sync for all da
devices that haven't seen final close already.

Ken
-- 
Kenneth Merry
ken@plutotech.com

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199810121936.NAA26254>